[PATCH] D26022: [X86][AVX512] Add patterns for all variants of VMOVSS/VMOVSD instructions

Igor Breger via llvm-commits llvm-commits at lists.llvm.org
Tue Nov 8 13:12:16 PST 2016


igorb added inline comments.


================
Comment at: lib/Target/X86/X86InstrAVX512.td:3324
+
+def : Pat<(v4f32 (X86Movss VR128X:$src0 ,(v4f32 (scalar_to_vector (f32 
+          (X86selects (i1 (trunc GR32:$mask)), (f32 FR32X:$src1), (f32 FR32X:$src2))))))),
----------------
please change indentation



================
Comment at: lib/Target/X86/X86InstrAVX512.td:3335
+
+def : Pat<(v2f64 (X86Movsd VR128X:$src0 ,(v2f64 (scalar_to_vector (f64 
+          (X86selects (i1 (trunc GR32:$mask)), (f64 FR64X:$src1), (f64 FR64X:$src2))))))),
----------------
The patterns for SS/SD is very similar, would you consider to implement it as multiclass ?


https://reviews.llvm.org/D26022





More information about the llvm-commits mailing list